You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Jan Hendrik <ja...@bigfoot.com> on 2004/04/07 15:31:57 UTC
experimental svn-win32-1.1-dev-9226: memory leak closed
Hi all out there!
Having been a long-time complainer about memory troubles (well,
by and large it got better through the months though meanwhile I
often let things go as they go or not go <g>) I'd like to post a very
positive report today:
I had noticed several times that slightly larger commits with
TortoiseSVN would hunt memory usage up through the roof
(300MB+) and crash even before TSVN showed any output in its
progress dialog while a commit with SVN commandline would use
about 15 MB only and succeed. Not being aware of the different
way TSVN and SVN commandline handle commits I finally posted
this problem to the TSVN list. Steve King then pointed me to the
same or similar commit through the commandline by doing
svn ci --target commit.txt.
Guess what? Memory went up through the roof again (SVN 1.0.1-
2.dev).
Second attempt, now with experimental build (thanks, Ben) svn-
win32-1.1-dev-9226.zip and commit succeeded, using a maximum
of 14 MB of memory!
Now I only wonder if it is wise to stay with the experimental build or
if that would/could likely cause other troubles, not least with either
or both Apache 2.0.48 (which I could update of course) and TSVN
1.0.3 built against SVN 1.0.1 and Apache 2.0.48.
Thanks for all the work and all the many, often invisible
improvements!
Jan Hendrik
---------------------------------------
Freedom quote:
Americans, indeed all free men, remember that in the final choice
a soldier's pack is not so heavy a burden as a prisoner's chains
-- Dwight D. Eisenhower
First Inaugural Address, January 20, 1953
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: experimental svn-win32-1.1-dev-9226: memory leak closed
Posted by Jan Hendrik <ja...@bigfoot.com>.
Concerning Re: experimental svn-win32-1.1-dev-
Ben Collins-Sussman wrote on 7 Apr 2004, 10:39, at least in part:
> On Wed, 2004-04-07 at 10:31, Jan Hendrik wrote:
>
> > Second attempt, now with experimental build (thanks, Ben) svn-
> > win32-1.1-dev-9226.zip and commit succeeded, using a maximum
> > of 14 MB of memory!
>
> Hm, I don't think r9032 (which I presume is the fix on /trunk which
> has solved the problem) has been backported to the 1.0.x branch.
> Perhaps it should be a 1.0.2 candidate? It looks like it could
> seriously affect tortoisesvn users...
That would be great indeed, Ben. It has troubled me several times
and a simple recursive commit from the commandline is not always
what I want. And this time I really wanted to keep a couple of files
out of the commit while splitting up would have been a bad thing
either. Though for the time being I'll go with the 1.1-build that would
not help TSVN ...
Jan Hendrik
---------------------------------------
Freedom quote:
The only justifiable purpose of political institutions
is to assure the unhindered development of the individual.
-- Albert Einstein
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: experimental svn-win32-1.1-dev-9226: memory leak closed
Posted by Ben Collins-Sussman <su...@collab.net>.
On Thu, 2004-04-08 at 11:38, Philip Martin wrote:
> How about this one?
Thanks, Philip. I've nominated both r8949 and r9032 for backport to
1.0.2.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: experimental svn-win32-1.1-dev-9226: memory leak closed
Posted by Philip Martin <ph...@codematters.co.uk>.
Ben Collins-Sussman <su...@collab.net> writes:
> On Wed, 2004-04-07 at 10:39, Ben Collins-Sussman wrote:
>> On Wed, 2004-04-07 at 10:31, Jan Hendrik wrote:
>>
>> > Second attempt, now with experimental build (thanks, Ben) svn-
>> > win32-1.1-dev-9226.zip and commit succeeded, using a maximum
>> > of 14 MB of memory!
>>
>> Hm, I don't think r9032 (which I presume is the fix on /trunk which has
>> solved the problem) has been backported to the 1.0.x branch. Perhaps it
>> should be a 1.0.2 candidate? It looks like it could seriously affect
>> tortoisesvn users...
>
> Hm, now I wonder. r9032 specifically fixed a runaway memory bug
> whenever 'svn up' tried to delete many things. But Jan's memory bug was
> about commits. So it must be some *other* memory-fix to /trunk that's
> at work here. Anyone know which one that is? That's the one we should
> backport.
How about this one?
------------------------------------------------------------------------
r8949 | julianfoad | 2004-03-09 15:43:49 +0000 (Tue, 09 Mar 2004) | 7 lines
Reduce memory usage of multiple-target commits (part of issue 1635).
Patch by D.J. Heap <dj...@shadyvale.net>.
* subversion/libsvn_client/commit.c
(svn_client_commit): Use a subpool in the target-processing loop.
------------------------------------------------------------------------
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: experimental svn-win32-1.1-dev-9226: memory leak closed
Posted by Ben Collins-Sussman <su...@collab.net>.
On Wed, 2004-04-07 at 10:39, Ben Collins-Sussman wrote:
> On Wed, 2004-04-07 at 10:31, Jan Hendrik wrote:
>
> > Second attempt, now with experimental build (thanks, Ben) svn-
> > win32-1.1-dev-9226.zip and commit succeeded, using a maximum
> > of 14 MB of memory!
>
> Hm, I don't think r9032 (which I presume is the fix on /trunk which has
> solved the problem) has been backported to the 1.0.x branch. Perhaps it
> should be a 1.0.2 candidate? It looks like it could seriously affect
> tortoisesvn users...
Hm, now I wonder. r9032 specifically fixed a runaway memory bug
whenever 'svn up' tried to delete many things. But Jan's memory bug was
about commits. So it must be some *other* memory-fix to /trunk that's
at work here. Anyone know which one that is? That's the one we should
backport.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: experimental svn-win32-1.1-dev-9226: memory leak closed
Posted by Ben Collins-Sussman <su...@collab.net>.
On Wed, 2004-04-07 at 10:31, Jan Hendrik wrote:
> Second attempt, now with experimental build (thanks, Ben) svn-
> win32-1.1-dev-9226.zip and commit succeeded, using a maximum
> of 14 MB of memory!
Hm, I don't think r9032 (which I presume is the fix on /trunk which has
solved the problem) has been backported to the 1.0.x branch. Perhaps it
should be a 1.0.2 candidate? It looks like it could seriously affect
tortoisesvn users...
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org